Part 3. Procompetitive Predicament: Why Does Google Pay Billions for the Default If Not to Foreclose Rivals?
Part 3 of 4 analyzing Judge Mehta's landmark US v. Google decision. Google argued it competed for the default position with Microsoft to the benefit of consumers. Judge Metha didn't buy it. What now?
Welcome back to Part 3 of our deep dive into the DOJ and States’ landmark victory in US v. Google.
As always, welcome to those of you who are joining us for the first time at Competition on the Merits! If you are enjoying this – please send to friends to subscribe, add yourself to the growing list of paid subscribers, or shoot me a note to let me know what you want to hear about. Or all of the above!
If you are not caught up with our first two entries in the US v. Google series – start here:
Causation Confusion: Why Judge Mehta Got the Causation Standard Wrong and Whether it Matters.
Foreclosure Fun & Games in US v. Google
Today we focus on Judge Mehta’s analysis of the various procompetitive efficiencies arguments Google raised at trial. But I think the best framing for this entry is a question that loomed over the entire trial: “If defaults are not important enough to foreclose competition, why did Google pay $26 billion a year for them?”
Let’s get into it and start with some economic theory and then get to business.
COMPETITION FOR THE MARKET
The great UCLA economist Harold Demsetz coined the term “competition for the field” to explore the relationship between market concentration and competition. The state-of-the-art method for evaluating that relationship 70 years ago and through the mid-1970s was the so-called “Structure-Conduct-Performance” paradigm. The SCP paradigm posited a direct positive relationship between market concentration and price – the fewer firms, the greater market concentration, the less competition. The fundamental insight was that we could measure the intensity of competition by counting the number of firms with our fingers.
Demsetz was famous for his critique of the SCP paradigm and others would empirically debunk it – though it returns from time to time in policy circles even without supporting evidence. For our purposes, Demsetz’s primary contribution was the observation that the relationship between market concentration and competition goes both ways. It is endogenous, as the economists say. This was – and remains – an incredibly important and influential insight in economic thinking. The fundamental idea is that while reducing the number of firms in a market could reflect a decline in competition, the same reduction could just as easily reflect the forces of competition at work. Some firms are able to operate a lower cost or produce a better product. As these firms grow in size, market concentration increases, and the number of firms falls, but average costs decrease or quality increases.
The policy payoff of Demsetz’s work was that we cannot just look out at the market, count the number of firms still standing, and say something meaningful about the intensity of competition. Demsetz pointed this out in “Why Regulate Utilities?” where he examined how “competition for the market” could generate competitive outcomes. Ex ante competition “for the market” might generate small numbers of firm – even one! – but competitive outcomes.
Think of the battlefield after an intense fight – or even after a game of “king of the hill.” There is one man standing atop the hill holding up the flag. One could look at the single person left standing and declare that there was not much competition – and the person enjoys a monopoly over top of the hill – but that would be a mistake! The casualties of war at the base and along the hill tell us counting the remaining firms is missing some important dimensions of the competition that took place.
The idea is a powerful one. And one that has been very influential both in industrial organization economics and antitrust law where naked reliance on market structure to infer harm to competition has certainly fallen out of favor in the courts over time – even if the new FTC and DOJ Merger Guidelines more directly embrace 1960s economics than their predecessors. But outside of its obvious relevance for merger policy, the concept of competition for the field can be expanded to help us understand how competition works in markets where suppliers compete for distribution – like US v. Google, where the key complaint is that Google’s default search contracts have harmed competition. Competition among search engines, say – between Google and Microsoft’s Bing (among others) – takes place on many dimensions. Search engines famously can be used at zero nominal price for users. So competition takes place on dimensions of quality – things like accuracy of search results, successfully getting users the answers they are looking for, latency, ease of use, and other features. But there is also competition to get the search engine in the hands of users – competition for distribution.
COMPETITION FOR DISTRIBUTION: OF SHELF SPACE AND CELL PHONES
Competition for the contract is incredibly common within our economy. It is ubiquitous and can take many forms. One of the best ways to think about competition for distribution is as the often unobserved competition that takes place before the competition for consumers. It might be the competition among soda brands to be the exclusive supplier in the sports stadium, or for eye level shelf space at the supermarket, or broadcast rights for your favorite TV series. Sometimes competition for the contract involves actual exclusive dealing. Sometimes it involves what economists often describe as “partial” exclusives, e.g. Coca-Cola pays for 80% of the soda shelf space.
Perhaps the most common form of competition for the contract is competition for shelf space. And that competition often involves partial exclusives rather than full exclusives —- this is because consumer demand for product variety is relatively high in supermarkets. The default contracts at issue in US v. Google are also “partial exclusives” in the sense that they do not demand true exclusivity from Apple or Android OEMs. So let’s stay with that familiar example for those of you who have read Part I & Part II of our series on US v. Google: shelf space contracts or “slotting contracts” as they are sometimes referred to.
There are two key economic questions to understand when it comes to partial exclusives that will help us understand the underlying economics here. The first is why do the contracts exist in the first place? Both Coca-Cola and the supermarket both want to sell soda. Coca-Cola could skip the contract and just allow the supermarket to decide how to stock the shelves on its own. Why do we see contracts that offer discounts and payments conditional upon the supermarket allocating a specific amount of share of shelf space to Coke?
The second key economic question is why shelf space contracts sometimes involve exclusivity provisions? That is, why do they include restrictions on the supermarkets ability to promote or display Pepsi? It is these restrictions that usually generate antitrust litigation. One obvious potential answer – and the very basis of DOJ’s theory of harm in the case – is that the function of the default contracts is to purchase Bing (and other search engine’s) exclusion. Or to be more precise, the function of the contracts is to prevent Bing from being able to compete for minimum efficient scale.
Google’s primary challenge in this case was to provide a better answer than that anticompetitive one to Judge Mehta. Obviously, Judge Mehta was not persuaded. The answers to both economic questions will help to understand the issues better and evaluate Google’s answer and Judge Mehta’s analysis.
For longer answers to the first question (“why do the contracts exist in the first place”), see Klein & Murphy (1988, and yes, that is Google’s economic expert Kevin Murphy) or my own work with Klein on shelf space contracts. I’ll run through a short answer to the first question and get quickly to second question which is at the heart of the antitrust issues here.
We can do a quick example to get the basic economics down. Imagine the economic incentive conflict facing Coke and supermarkets over how the supermarket allocates its prominent shelf space. The supermarket is not some passive transmitter of consumer preferences – how it shelves products, what it promotes, and which products it shelves matter for consumer behavior! For example, marketing studies suggest eye-level shelf space or other prominent display can increase demand for the promoted product by 2-4%. That may not sound like a lot! But selling 2% more is a big deal to Coke. Do you know the marginal cost of a can of Coke? It is very low relative to the wholesale price at which it sells to the supermarket. Those incremental sales induced by the extra promotion from the supermarket (in this case, eye-level shelf space) are worth a lot.
But it is not worth as much to the supermarket for two reasons. The first is that more sales of Coke means fewer sales of Pepsi for the supermarket. The supermarket’s margin on Coke or Pepsi is about the same and shifting from one to the other does not really matter much from the retailer’s perspective. So giving Coke the promotional sales might help on the Coke sales but cannibalize Pepsi sales. The second is that while Coke’s margin on incremental sales is really high, the supermarket’s incremental profit margin on additional Coke sales is pretty darned low. Supermarket margins are generally in the 1-3% range.
In short – the incremental value to Coke to promotional shelf space is high; the value is relatively low to the supermarket. We get one exactly what expects where there are gains from trade! Coke contracts with retailers for the promotional shelf space. Or rather, the supermarket holds a competition and competing soda brands compete for those allocations with some form of payment (lump sum, a reduction in the wholesale price, etc.).
OK, so that is the answer to question number 1. And that helps to explain why we see contracts like Coke’s (and Google’s) emerging from competition for distribution. Shelf space (in the supermarket or on the iPhone) is valuable! And we can expect Apple or the supermarket to economize on it.
The second question – why we often see exclusives or partial exclusives with those contracts – gets us closer to antitrust issues. For a longer and more formal version of this explanation, see Klein and Murphy (2008). But the basic idea is relatively simple. Keeping in line with our supermarket example – imagine a world where retailers face downward sloping demand and get to make some choices about what to stock, how to stock it, and what to promote without losing all of their sales. This describes all retail markets so is not much of an assumption. Retailers can shift sales around by promoting one brand over another – whether that is by giving eye level shelf space, all of the shelf space, or some other form of preference. Assume some consumers are very brand loyal – the Coke consumer will buy Coke wherever you put it on the shelf – and some consumers are not brand loyal and are willing to switch around and use whatever is on promotion.
You can think of the supermarket as being able to deliver to one brand (Coke, Pepsi, Google, Bing) all of its non-brand loyal sales. This does not work if consumers will switch entire stores because Pepsi is not in the store or the customer has to reach to the bottom shelf to get it – unlikely to be the case in the supermarket. But the supermarket knows it has a valuable asset of being able to deliver those incremental, non-loyal consumers to a manufacturer within a product category and so “sells” those sales to a competing manufacturer. Perhaps it auctions the eye level shelf space to Coke or Pepsi. Or Apple auctions its default space off to Google or Bing. It is a method for the distributor to deliver valuable promotional sales.
This helps answer why we see exclusivity or partial exclusivity. The supermarket offers exclusive or partially exclusive access to it’s customers for Coke and in exchange gets compensated in the form of lower wholesale prices or other remuneration. Obviously, Coke has to pay the supermarket more than it would lose from any losses it sustains from consumers who switch stores because the supermarket does not carry Pepsi.
Exclusivity – even over the promotional sales – shifts the competition between Pepsi and Coke from ex post to ex ante. That is an important point because ex ante competition often has a greater sales response than ex post competition. Think of ex post competition as walking into the supermarket and deciding between a can of Coke or a can of Pepsi. When Coke reduces the price of its product there is some switching from non-brand loyal Pepsi drinkers. The loyal Pepsi drinkers continue to drink their preferred brand. But when competition is moved ex ante, you can think of the retailer as choosing between all competing manufacturers for all of its sales. Each of the brands becomes a very close substitute and the winner of the auction for the exclusive sees a huge increase in its sales. From an elasticity of demand perspective you can think of the competition for exclusive making each manufacturer’s demand more elastic and thus generating a lower wholesale price. This is good from a competition perspective!
So why do we see contracts for shelf space where distribution choices are important in influencing consumer purchases? Because supplier margins are much larger than retail margins and so they have an incentive conflict with respect to promotional effort that can be resolved by contract.
Why exclusives? Because the shelf space is much more valuable to the distributor (supermarket, iPhone) when it is offered on an exclusive basis. The exclusivity or partial exclusivity allows the distributor to “sell” the non-loyal customers to the supplier – thus increasing the value of his shelf space.
What potential consumer benefits derive from this sort of competition for the contract? You can think of two different types of potential answers.
The first is that competition between Coke and Pepsi might generate lower soda wholesale prices and thus lower retail prices for consumers.
The second is that large payments from Coke and Pepsi to supermarkets might be passed through to consumers not in the form of lower soda prices but on other competitive dimensions (perhaps an extra check out aisle, a butcher, add an ATM, etc.).
I do not want to spend much time on the law today. But suffice it to say that the courts have uniformly accepted “competition for the contract” as a legitimate form of competition and one that deserves protection under the Sherman Act. These sorts of benefits are often recognized in exclusive dealing cases. No court disputes that. Judge Mehta certainly did not. In fact, we can think of the entire Google case as a debate about whether default contracts promote or suppress competition for the contract.
And of course, these potential answers are competing against the anticompetitive hypothesis that forms the foundation of DOJ’s complaint: that the contracts are designed to foreclose rivals from sufficient distribution to reach minimum efficient scale. We dealt at great length with the underlying economics of foreclosure in our last entry. So please refer to that if interested.
But the foreclosure theory is important to keep in the back of your mind here – because procompetitive explanations and anticompetitive ones are intimately intertwined in antitrust analysis. Persuading the court on a positive explanation for your conduct often leads courts to be skeptical of anticompetitive explanations; and vice versa.
We’ve got all the economics we need now to dive into Judge Mehta’s analysis.
DOES JUDGE MEHTA’S COMPETITION FOR THE CONTRACT ANALYSIS GIVE AWAY THE FARM ON CAUSATION?
The discussion of competition for the contract in US v. Google is a little chaotic. Or perhaps disorganized is the right word. Where does consideration of competition for the contract arguments go within the antitrust framework?
Recall the Microsoft framework:
From a century of case law on monopolization under § 2, however, several principles do emerge. First, to be condemned as exclusionary, a monopolist's act must have an “anticompetitive effect.” That is, it must harm the competitive process and thereby harm consumers. In contrast, harm to one or more competitors will not suffice.
Second, the plaintiff, on whom the burden of proof of course rests, must demonstrate that the monopolist's conduct indeed has the requisite anticompetitive effect. In a case brought by a private plaintiff, the plaintiff must show that its injury is “of the type that the statute was intended to forestall,” no less in a case brought by the Government, it must demonstrate that the monopolist's conduct harmed competition, not just a competitor.
Third, if a plaintiff successfully establishes a prima facie case under § 2 by demonstrating anticompetitive effect, then the monopolist may proffer a “procompetitive justification” for its conduct. If the monopolist asserts a procompetitive justification — a nonpretextual claim that its conduct is indeed a form of competition on the merits because it involves, for example, greater efficiency or enhanced consumer appeal — then the burden shifts back to the plaintiff to rebut that claim.
Generally, one can think of Google’s positive explanations for how its default contracts enhance competition as falling into the third category, i.e. “procompetitive justifications,” that come into play only if the plaintiff has established its prima facie case.
Judge Mehta’s first discussion of competition for the contract begins with a different framing. Responding specifically to Google’s arguments that it wins the default arrangements because it is a superior product – and so its behavior cannot possibly be exclusionary under Section 2 of the Sherman Act – Judge Mehta gives some and takes some.
Judge Mehta concedes that Google has often won competitions for the default because it is the superior product; but adds that perhaps its superior quality is in turn durable precisely because it is a monopolist. I say that this is a somewhat odd home for these points analytically for a few reasons. The most important is that Judge Mehta bites off more than he needs to in framing the issue – and in so doing highlights a weakness in DOJ’s prima facie case and his analysis of it.
So, there is competition for the contract, and Google has indeed prevailed at least in part because of its superior quality. Here is where Judge Mehta takes an odd turn:
The danger in saying that there is no real competition for the contract is that if Google does not have a real constraint on its monopoly power in search and is not responding to a real competitive threat, its conduct is not impacting competition. It may well harm rivals, but if the court concedes that rivals were never going to succeed because Google acquired its monopoly power through lawful conduct and innovation, we return straight to the Rambus-causation problem we discussed in Part II.
I do not think Judge Mehta needed to go this far. And I just want to highlight for now that it creates a causation vulnerability in the opinion I suspect Google will highlight on appeal. Here is perhaps the most aggressive passage that reads quite a bit like “Google would win the competition for contract no matter what.”
But I do not think this passage does the work Judge Mehta thinks it does. Too the contrary, if Google would win the competition for contract no matter what – and under any counterfactual – it is impossible to conclude that Google’s conduct is the but-for cause of its monopoly position.
Greg Werden makes this point eloquently:
The court held that the exclusionary conduct must be “reasonably capable of contributing significantly to a defendant’s continued monopoly power.” Thus, the government had to prove that the threats to Microsoft’s operating-system monopoly were real, but not that they likely would have succeeded. The government’s burden is similar in Google.
The government’s proposed conclusions of law in Google quote the Grinnell elements. And its post-trial brief asserts that the conduct complained of “allowed Google to maintain its monopoly power in violation of Section 2 of the Sherman Act.” But the government does not contend that rival search engines ever posed a real threat to Google’s monopoly. Indeed, it claims to have proved just the opposite.
This portion of the opinion gets bogged down in the question of whether Google’s default contracts can be exclusionary. Exclusive dealing contracts from a monopolist that substantially foreclosure rivals and have an anticompetitive effect are not competition on the merits. Most of this was superfluous. Judge Mehta could have jumped immediately to that part of the analysis. But his extraneous finding that Google would have won that competition no matter what creates some vulnerability on causation. But I want to move beyond causation and get to Google’s rebuttal case, i.e. the procompetitive explanations Google offers for its default contracts.
DO THE GOOGLE DEFAULTS HAVE PROCOMPETITIVE EFFECTS?
Recall from the Microsoft framework, the burden lies with Google here to demonstrate procompetitive effects of its conduct that are non pretextual. It offers a few but I want to focus on some obvious ones. Recall from our shelf space example that one benefit of Coke competing with Pepsi for eye level shelf space is that competition at the supermarket level means those payments will be passed on to consumers on some margin.
Google similarly argues that its payments generate lower device prices and are passed on to consumers on other margins. This testimony largely comes in through economic experts. Here is Murphy on this point at trial:
The basic argument is that while default payments to Apple skyrocketed, the device margins remained the same, implying substantial pass-through of those payments. Notice that I said implying, not demonstrating. My own review of the evidence is that Google explained with economic logic that the simultaneous increase in payments for distribution and constant device margins implies pass-through to consumers. This is the analytical equivalent of the point that Coke’s large payments to retailers while retailers maintain constant margins implies pass-through to consumers. It may not be in the form of a lower price of soda. Retailers might choose to use that money to compete on other margins that attract consumers from other stores. The logic is sound enough. In fact, I agree it is sound.
But other than point to the correlation of the increase in defaults and constant margins, and explaining the implications for pass-through, it does not demonstrate any causal nexus between the payments and lower device prices.
But does it meet Google’s burden? That is the $64,000 question, isn’t it? First, Judge Mehta’s analysis and then a few observations. Judge Mehta underscores that Google’s expert does not show conclusively the relationship between default payments and other benefits, be they lower device prices or increased search output:
Here is Judge Mehta with the same point – the relationship is not demonstrated conclusively nor is pass through contractually obligated:
And again with respect to the argument that the payments reduce Android prices, Judge Mehta points out that causality is not established:
It’s time for a few observations:
First, back to the law – recall that the Microsoft framework does not require a causal relationship to be established between the procompetitive effects and the underlying conduct. Google’s burden at this stage was merely to show non-pretextual explanations for its conduct. Did it do that? Judge Mehta appears to concede the arguments are plausible and not pretextual. But they fail because they are not quantified or shown at a causal level. I do not think Section 2 law requires as much to shift the burden back to the plaintiff.
Second, it is quite an odd juxtaposition that Judge Mehta embraces whole heartedly the reduced causation standard when it comes to plaintiffs’ prima facie burden (erroneously, as we have explained) but appears to hold the defendant to a higher causation standard in its rebuttal case. Again and again, Judge Mehta finds that Google fails to prove its procompetitive explanations because it fails to establish what would have happened with respect to its claimed benefits in the absence of the exclusive defaults. But Judge Mehta excuses the plaintiff from confronting the same counterfactual question when it comes to foreclosure and causation. He cannot be right both times.
Third, odd enough for reversal? I’m not sure. I doubt it on its own. I do think Judge Mehta’s causation error infects the efficiencies analysis. And certainly highlights that error. But I do not think the quality of evidence presented on procompetitive efficiencies is overwhelming. That is not surprising for a reason captured in the economics primer above. Like Coke’s payments to supermarkets can be passed on to consumers on just about any competitive margin, so it is with Google’s default payments to Apple and Android OEMs. Perhaps that is research and development, perhaps is is device promotion, perhaps device prices. But it is not surprising that it is easier to establish pass-through as a matter of economic logic than it is as a matter of causal identification.
SO WHY DID GOOGLE ENTER THESE DEFAULT CONTRACTS, ANYWAY?
So back to the bottom line. To win at trial, Google was always going to have to provide a winning answer to the question: “Why did you pay billions of dollars for search defaults if not to maintain monopoly?” All kinds of silly answers have been thrown around. Was it to keep Apple out of the search market? (There is not an ounce of evidence and not even DOJ would allege as much)
Google would need to come up with a convincing and persuasive answer to the question of why it would pay $26 billion for search defaults if not foreclosure of rivals. Clearly it did not persuade Judge Mehta. But what is the answer?
I think the answer goes back to the economics of Coke and shelf space. It is difficult to understand just how valuable those incremental, promotional “searches” are to a search engine. They drive ad revenue. They improve quality. These are all good things. But imagine a world in which brand loyal search users will switch from the default but the rest will use the default provided. Like the supermarket owns the valuable shelf space to commit to deliver the marginal consumers to Coke or Pepsi – whichever bids the most; Apple and Android OEMS own the valuable real estate here. Apple and Android OEMs can deliver incredibly valuable marginal searchers to either Google or Bing.
The Court seemed persuaded that Apple’s real estate was incredibly valuable. It even seemed persuaded that there were real auctions and competition for that shelf space. But Judge Mehta could not get past the fact that Google won those competitions. What I find interesting about Judge Mehta’s conclusion that the competition was distorted or somehow not “on the merits” because Google won most of the time is his own findings that Google had superior quality and lower costs.
It is not a surprising outcome – even with intense competition – that the higher quality and lower cost provider wins most of the time and earns more market share. Not surprising at all. Murphy tries to address the idea that the key question is whether the rival can compete not whether it wins:
And that is where I think Judge Mehta got off of Dr. Murphy’s bus. He could not envision a world where losing competition constrains the winner. In other words, Judge Mehta looked at Google standing atop the hill and concluded that the competition to get there was not meaningful. The classic fallacy Demsetz warned of. Now to what end? Again, Google made out a strong enough case to establish these procompetitive efficiencies as non-pretextual. But one has the distinct feeling that Judge Mehta would have then ruled for DOJ in the “balancing” stage anyway.
So how big of a deal is this issue on appeal? It matters for aesthetic reasons. Judges want to understand the “why.” And the more comfortable they are understanding why and what these contracts are doing the better. I suspect Google briefs these arguments aggressively – highlighting the asymmetric causation standards Judge Mehta applied to DOJ and Google as well as the logic that the payments must have been passed on as a matter of basic economic theory. But if the decision is reversed on appeal, I very much expect the action to be surrounding skepticism of Judge Mehta’s causation and foreclosure analysis.
That will do it for Part 3! Next time in Part 4 we will talk about Remedies!